ข่าว

I’ve been locked in to my various screens since results from midterm elections started to roll in last night. And it didn’t take long to see the beginning of the blue wave: Democrats took control of both the US House of Representatives and the Senate, a dramatic swing from Donald Trump’s victory two years ago.   While the results are striking, we probably won’t be seeing some of the big climate and energy policy shifts that have come in recent years as Congress and the presidency have changed hands. Instead, we’re entering a period of divided government, with a Congress that will likely do battle with Trump’s White House over the next two years. But even if the federal government isn’t likely to pass laws that overhaul energy and climate policy, the midterm results could be a big deal for energy, especially at the state and local level.  Let’s step back for just a moment to put these results in context. Readers might remember huge laws affecting climate and energy that were passed in just the last four years. The 2022 Inflation Reduction Act included massive support for clean energy, including tax credits and government grants and loans.Then the 2025 One Big Beautiful Bill Act essentially slashed much of the support and financial incentives in the IRA from two years before. What those two laws had in common was a period when the government was entirely united, with the Senate, House of Representatives, and presidency all controlled by the same party—the Democrats in the first case and the Republicans in the second Now Democrats have captured a majority in both the Senate and the House. But Donald Trump is still in the White House. And crucially, neither chamber is anywhere close to a two-thirds Democratic majority. That’s the ratio that would be needed to override a presidential veto. So even if Congress puts together new legislation to boost climate measures and clean energy, Trump will need to sign it—and that dramatically narrows the scope of what might happen legislatively between now and the next election. There could be some bipartisan compromises. One interesting bit of legislation to watch immediately after the election is what happens with permitting reform. In late September, a bipartisan group of senators put forward a bill that would overhaul how new energy projects are reviewed and approved in the US. The law aims to cut red tape and speed up the permitting process. The Senate is set to vote on that package, called the Bipartisan American Affordability and Jobs Act, on November 9, before the new Congress takes control in January. The White House has supported the bill, and it could pass the Senate if just a handful of Democrats join the Republican majority. But now, we might see Democrats wait until next year to try to renegotiate a deal, though anything would still need to pass muster with Trump. With Democratic control of Congress, we’ll likely see much more scrutiny of Trump and agencies controlled by his appointees. Democratic leaders have already signaled plans to go after leaders like EPA commissioner Lee Zeldin over cuts to staff and investments. Democrats will have the power to subpoena cabinet members and open investigations. Looking beyond the national results, a major trend to watch moving forward will be local action. Democrats swept local control of Pennsylvania, and officials there have signaled plans to support renewable energy and efforts to cut greenhouse-gas emissions. Democrats also took control of the Georgia Public Service Commission for the first time in 30 years. The group regulates electric utilities, essentially controlling much of the grid’s buildout.  Among the broadest effects of these election results, though, will be for the data center industry. According to tracking from Ballotpedia, there were 41 local ballot questions related to data centers. And overwhelmingly, voters chose to restrict their development. Nearly every single one of these measures looks set to come down against data center development, as of the morning of November 4.  And AI critics fared well in their races, too. Democrat Gina Hinojosa won the governor’s seat in Texas, ousting Republican incumbent Greg Abbott. During her campaign, Hinojosa called for a law to completely pause data center development; if passed, it could have major implications for both the state’s grid and the data center industry more broadly. We might not see dramatic new climate- or energy-related federal legislation between now and the next major US election. But many voters signaled, and quite loudly, that they wanted a change. We’ll see how much will shift in the next two years, and where that leaves us.  This article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here. 

Texas is the fastest-growing data center market in the US and ground zero for Big Tech’s largest AI buildouts. But the blue wave that swept the Lone Star State may jeopardize that position. Democrat James Talarico won the US Senate seat against Ken Paxton. Gina Hinojosa won the race for governor against the Republican incumbent, Greg Abbott. At the state level, Democrats will take control of the Texas House for the first time in 24 years. Talarico and Hinojosa made data centers a central issue, discussing them constantly on the campaign trail. Even before the polls closed, insiders in the data center industry noticed that the times were changing. At the industry’s DCD conference in Ashburn, Virginia, on Tuesday, panelists lamented that Texas had been the easiest place to build for years—until September. That’s when Abbott, perhaps seeing the increasingly bipartisan distaste for these facilities, began changing course himself.  Developers said the governor’s office had stopped responding to their emails. On September 21, Abbott directed a key state regulator to halt data center permits. Many panelists at the conference were expecting this to mean that existing projects might be slowed and new ones could be halted altogether.  The governor-elect indeed seems eager to do that. Hinojosa supported tax breaks for data centers in 2023 as a way to boost construction jobs, but during her campaign she began pushing for a full-fledged moratorium. That would be a problem for the hyperscalers of Big Tech, which have ambitious plans for AI training facilities in the state. The scale of projects already underway in Texas is massive. The Stargate Frontier data center in Shackelford County is projected to serve 1,820 megawatts—about as much as the nation’s largest data center, SpaceX’s Colossus 2 in Tennessee, and more than twice as much as the largest data center currently operating in Texas. Nexus Hubbard, also under construction and intended mostly to serve Anthropic, has a planned size of 6,400 megawatts and is the second-largest data center being built in the US, behind only Meta’s Hyperion project in Louisiana.  The potential for Democrats to order delays to these projects or moratoriums on new ones adds to growing anxiety that the AI buildout could stall. Banks and lenders are already getting pickier about the projects they back. And given the extent to which the data center buildout is underpinning growth in the economy right now, those forces combined could have ripple effects. Voters around the country, though, were unambiguous in the midterms, sending a similar message on everything from state ballot questions to national offices. Voters supported efforts to protect ratepayers from increases in their electricity bills caused by data centers, and they favored some larger bans on new data centers altogether. If we needed any more convincing that Big Tech has fumbled its effort to sell voters on the idea that AI-powered abundance will make health care better, goods cheaper, and scientific breakthroughs more common, this was it. A key question is whether the Texas counties with the largest data center presence were part of the reason these Democrats got elected, or if the anti-data-center vote came from elsewhere.  It’s complicated: The counties where much of the data center buildout is happening are around Dallas. Dallas County and others closest to it predictably went blue, but nearby Ellis County, which leads in permitting requests for data centers, went for Greg Abbott and Ken Paxton. Cameron County, near the state’s southern border, is a hot spot for data centers and voted blue in the midterms, which was a shift. But Mitchell County and Scurry County, where a bunch of projects in west Texas are located, went red.  This is a big deal, because no single office in the state decides whether a project goes ahead or is halted. Decisions often reflect an interplay between city governments, county officials, and state offices.  Local officials who have some say over data center development may resist sweeping restrictions from the state. And Democrats pushing for a moratorium might also face pushback from the construction unions that supported some of their campaigns. So while the blue wave spells a death knell for Texas data centers in theory, election winners may find that following through on their opposition once in office is much more complicated than campaigning on it.

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. At the end of September, President Trump signed an executive order that contained just one directive, ordering government agencies to rebrand what they had been calling artificial intelligence as “Super Intelligence.” That term beat out other contenders, including “Extreme Intelligence” and “Supreme Intelligence,” offered by the president in a poll on X a couple of weeks earlier.  Rebrands are, of course, now a White House tradition: Trump has changed names to project US dominance of certain bodies of water (the Gulf of America and Lake America), to signal new institutional priorities (Department of War), or simply to leave his mark on storied places (see the ongoing fight over renaming the Kennedy Center).  The stated reason for this name change is that today’s AI systems “amplify human ingenuity” rather than merely “imitate” intelligence, and the government’s terminology should reflect that. But if this rebrand follows the pattern of others, its effects will extend beyond government.  It will turn a choice of words into a political loyalty test: Those who use the new term demonstrate allegiance to the administration, and those who don’t become marks for attacks (recall that the Trump administration barred the Associated Press from some White House press events after the news outlet did not adopt the term “Gulf of America”).  But most Americans go months without ever pondering what to call the major lake between New York and Canada. AI, on the other hand, is at the center of our national conversation and a technology two-thirds of Americans report interacting with multiple times a week. It might be the boldest rebrand yet.  So one month in, who’s on board? Let’s start with the government itself. The term “Super Intelligence”—or simply “SI”—now appears more often on government solicitations, like a new request by Immigration and Customs Enforcement for companies to build “SI-powered” surveillance for ICE detention facilities. The new term has been swapped in silently on most government websites. Some, including that of the National Institutes of Health, now include a banner warning visitors they might see both the new and the old terms during the transition period. There have been hiccups. A day after the rebrand order, the Pentagon announced an ambitious new initiative to study the future of warfare, called Project Meridian; it used the old terminology. A Federal Trade Commission press release last week used “superintelligence” instead of the two-word, capitalized version ordered by the president, and recent press releases from Health and Human Services show no sign of the new language.  The term “AI” will likely stick around given the countless AI-related offices, initiatives, and programs that have sprung up in recent years, like the Chief Digital and Artificial Intelligence Office at the Pentagon (the executive order says it doesn’t apply to previously issued regulations or documents). As for government employees, it appears to be taking time for “SI” rather than “AI” to roll off the tongue. One researcher at a national lab I interviewed a few weeks ago used “AI” repeatedly before sheepishly acknowledging the slip-up. Even Defense Secretary Pete Hegseth used “AI” in an interview on October 3.  But Trump’s other rebrands suggest the goal isn’t merely to change the government’s vocabulary. It’s to change how the rest of us speak, too. In the past couple of weeks, most of President Trump’s posts on Truth Social have been endorsements for candidates in the midterms. But on October 27 he took a break from electioneering to criticize the Wall Street Journal for not using “Super Intelligence,” “the name which virtually everyone else is using, or soon will be.” He suggested that a new rule will soon go further than the executive order to outright ban “artificial intelligence” from government documents.  But the Wall Street Journal is not alone: Even Newsmax, an outlet whose founder described it as “supportive of the President and his policies,” isn’t using the president’s preferred term.   Are Big Tech executives falling in line? It’s complicated. The same day that President Trump signed the rebranding order, he also released a nonbinding statement on AI safety, the White House Accord on Super Intelligence, signed by six of the most powerful people in tech: Sundar Pichai of Google, Dario Amodei of Anthropic, Mark Zuckerberg of Meta, Greg Brockman of OpenAI, Elon Musk of xAI, and Jensen Huang of Nvidia. But even among this group—three members of which attended Trump’s inauguration—the term has not caught on.  In an analysis of third-quarter earnings calls, CNBC reported that only Zuckerberg and Musk used the term consistently. But Zuckerberg doesn’t quite count: Meta has been saying “superintelligence” since last year anyway. Musk’s companies seem to use the term the most—and he even pledged to rename SpaceXAI to SpaceXSI. Musk himself still mixes SI and AI when posting on X.  Executives from Amazon, Microsoft, and Alphabet used “AI” in their calls, though Microsoft’s recent letter to shareholders referred to the “Super Intelligence era.” Huang has referenced the term but still uses AI too. One objection to this rebranding project is that it will cost time and taxpayer dollars. It is Salesforce CEO Marc Benioff’s prerogative to decide that SIForce is a better name than AIForce and direct his company to implement that change. But government employees are now spending time writing guides on how departments should navigate the rebrand. And, per the executive order, leaders of federal agencies will be required to put their heads together to draft a new legislative definition of “Super Intelligence” to be run by the president by the end of this month.  But I’m more worried about what this does to our national conversation about AI than I am about the potential waste of time and money. Surprisingly often I overhear strangers at a grocery store or coffee shop who have struck up a conversation about AI. The substance of the conversation is rarely uplifting—they most often talk about

A new approach known as “reverse game theory” aims to control powerful AI models by designing systems that incentivize them to stick to the rules. As AI models become more capable, safety researchers are scrambling to ensure they are “aligned” with human desires—to make sure, in other words, that rogue agents don’t break the law, hack into unauthorized servers, or release poison gas into major cities and kill us all. Until now, this work has mostly focused on trying to peer into an AI’s “brain” to understand what it is “thinking,” or training models to hew to values that we hold dear. Claude, for instance, has its constitution—a sort of moral code. The trouble is, these techniques don’t seem to work particularly well: Models have been shown to deliberately underperform on tests, hide their intentions, and mislead their evaluators.  Now some alignment researchers believe that reverse game theory, also known as mechanism design, might get better results. Where game theory tries to predict how rational actors will behave when given a particular set of rules, reverse game theory attempts to design rules that will elicit the kinds of behavior we want to encourage. An example of mechanism design in the real world might be the classic strategy for splitting a piece of cake in two: One person cuts the cake and the other gets the first choice of pieces. This incentivizes the cutter to make the pieces as even as possible. Other examples include surge pricing in ride-hailing apps, which encourages drivers to be available at busy times, and second-price auctions, where the winning bidder pays the price of the second-highest bid instead of the highest. Platforms use these auctions to sell online advertising because they encourage buyers to bid close to the maximum they are willing to pay. In AI, mechanism design starts from the premise that we may never be able to understand enough about the internal workings of frontier models to make sure they are aligned. This means we should design mechanisms that incentivize agents to behave in ways that we want, regardless of what’s actually going on under the hood.  A recent preprint paper by Stephen Morris, an economist at MIT, along with colleagues from Yale and Columbia suggested some potential mechanisms that could stop agents from “sandbagging” tests (pretending to be worse than they actually are—for example, to avoid triggering safety mechanisms).  Instead of being rewarded only for how well it does on a test, a model could be rewarded for how well its answers predict what a second model will say. Deliberately playing dumb becomes a losing bet. “AI agents have a very good understanding of what other AI agents can achieve,” Morris says. “So it does seem like a context where this cross-checking could work more practically.  The paper also suggests pitting agents against each other, creating a competitive dynamic that incentivizes them to perform their best, or pairing agents that are known to be biased in opposite directions and rewarding them depending on how closely their outputs match, thus incentivizing them to reach an optimal answer.  A second approach, as envisioned by Elad Hazan, a professor of computer science at Princeton, is to design mechanisms that take inspiration from the ways human society keeps people “aligned”: social norms, the justice system, the monetary incentives of free-market capitalism. “We want to align superintelligent AI,” he says. “Why don’t we first talk about humans? How do we align humans?” Hazan’s suggested mechanism involves having AI models take turns as either “solvers” that tackle problems or “auditors” that check the others’ work. “One agent acts like law enforcement; the other agent acts like the citizen,” he says.  During training, the models are rewarded depending on outcomes: Solvers that get caught cheating are punished, but so are auditors that fail to catch the cheaters. In situations where there’s a disagreement between auditors and solvers, the case can be sent to “trial”—perhaps a human expert will decide who is right.  There is, Hazan admits, a risk that agents will collude with each other—as auditors have colluded with the businesses they’re meant to be auditing, or corrupt police officers have colluded with criminals. But that problem can be solved by rewarding the auditors well enough, so that being caught isn’t worth the risk (a lot easier in the world of AI training than the real one, where money for police salaries is tight), and by auditing the auditors using techniques akin to employing mystery shoppers.  Hazan points to Nobel Prize–winning work by the economist Gary Becker, who introduced the idea of the “rational offender” in the 1960s. Criminals don’t break the law because they’re inherently bad people, he said. Instead, they do it when it’s the most rational choice given the incentives on offer. Crime makes sense only when the proceeds outweigh the potential punishment.  Alignment shouldn’t be thought of as a property of a model, he argues, but as a descriptor of its behavior at a particular moment. “You are, like, a nice person,” he says. “Are you aligned? I don’t know. Probably you are aligned most of the time, but there could be a situation where you’re not.” One potential issue with applying mechanism design to AI: It relies on the assumption that AI agents behave rationally. Mechanism design has a mixed record with humans, who often fail to behave the way economic models say they should.  It’s ironic: Economists have spent decades trying to reduce human behavior to mathematics, and now, in large language models, we’ve built mathematical systems that run on human language and could be just as inscrutable. AI models exhibit behaviors that seem more human than machine, and they can be swayed by subtle factors that are difficult to understand and explain. “Is it even reasonable to assume that they will respond to incentives?” asks Helen Qu, a research fellow at the Flatiron Institute, a nonprofit dedicated to advancing scientific discovery through computational methods, who has written about mechanism design and